Upload 3 files
Browse files- checkpoint.pkl +3 -0
- config.yml +145 -0
- stats.h5 +3 -0
checkpoint.pkl
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:2cb1734dad1520243cb47d26b77f13cc63c221fadceb0436b3dde33f32fd759c
|
3 |
+
size 104079309
|
config.yml
ADDED
@@ -0,0 +1,145 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
allow_cache: true
|
2 |
+
batch_max_steps: 8100
|
3 |
+
batch_size: 64
|
4 |
+
config: conf/multi_band_melgan.v2.yaml
|
5 |
+
dev_dumpdir: dump/dev_all/norm
|
6 |
+
dev_feats_scp: null
|
7 |
+
dev_segments: null
|
8 |
+
dev_wav_scp: null
|
9 |
+
discriminator_grad_norm: -1
|
10 |
+
discriminator_optimizer_params:
|
11 |
+
amsgrad: true
|
12 |
+
eps: 1.0e-07
|
13 |
+
lr: 0.001
|
14 |
+
weight_decay: 0.0
|
15 |
+
discriminator_optimizer_type: Adam
|
16 |
+
discriminator_params:
|
17 |
+
channels: 16
|
18 |
+
downsample_pooling: AvgPool1d
|
19 |
+
downsample_pooling_params:
|
20 |
+
count_include_pad: false
|
21 |
+
kernel_size: 4
|
22 |
+
padding: 1
|
23 |
+
stride: 2
|
24 |
+
downsample_scales:
|
25 |
+
- 4
|
26 |
+
- 4
|
27 |
+
- 4
|
28 |
+
in_channels: 1
|
29 |
+
kernel_sizes:
|
30 |
+
- 5
|
31 |
+
- 3
|
32 |
+
max_downsample_channels: 512
|
33 |
+
nonlinear_activation: LeakyReLU
|
34 |
+
nonlinear_activation_params:
|
35 |
+
negative_slope: 0.2
|
36 |
+
out_channels: 1
|
37 |
+
scales: 3
|
38 |
+
use_weight_norm: true
|
39 |
+
discriminator_scheduler_params:
|
40 |
+
gamma: 0.5
|
41 |
+
milestones:
|
42 |
+
- 100000
|
43 |
+
- 200000
|
44 |
+
- 300000
|
45 |
+
- 400000
|
46 |
+
- 500000
|
47 |
+
- 600000
|
48 |
+
discriminator_scheduler_type: MultiStepLR
|
49 |
+
discriminator_train_start_steps: 200000
|
50 |
+
discriminator_type: MelGANMultiScaleDiscriminator
|
51 |
+
distributed: false
|
52 |
+
eval_interval_steps: 1000
|
53 |
+
fft_size: 2048
|
54 |
+
fmax: 7600
|
55 |
+
fmin: 80
|
56 |
+
format: hdf5
|
57 |
+
generator_grad_norm: -1
|
58 |
+
generator_optimizer_params:
|
59 |
+
amsgrad: true
|
60 |
+
eps: 1.0e-07
|
61 |
+
lr: 0.001
|
62 |
+
weight_decay: 0.0
|
63 |
+
generator_optimizer_type: Adam
|
64 |
+
generator_params:
|
65 |
+
channels: 384
|
66 |
+
in_channels: 80
|
67 |
+
kernel_size: 7
|
68 |
+
out_channels: 4
|
69 |
+
stack_kernel_size: 3
|
70 |
+
stacks: 4
|
71 |
+
upsample_scales:
|
72 |
+
- 5
|
73 |
+
- 5
|
74 |
+
- 3
|
75 |
+
use_causal_conv: false
|
76 |
+
use_weight_norm: true
|
77 |
+
generator_scheduler_params:
|
78 |
+
gamma: 0.5
|
79 |
+
milestones:
|
80 |
+
- 100000
|
81 |
+
- 200000
|
82 |
+
- 300000
|
83 |
+
- 400000
|
84 |
+
- 500000
|
85 |
+
- 600000
|
86 |
+
generator_scheduler_type: MultiStepLR
|
87 |
+
generator_type: MelGANGenerator
|
88 |
+
global_gain_scale: 1.0
|
89 |
+
hop_size: 300
|
90 |
+
lambda_adv: 2.5
|
91 |
+
log_interval_steps: 1000
|
92 |
+
num_mels: 80
|
93 |
+
num_save_intermediate_results: 4
|
94 |
+
num_workers: 4
|
95 |
+
outdir: exp/train_nodev_all_vctk_multi_band_melgan.v2
|
96 |
+
pin_memory: true
|
97 |
+
pretrain: ''
|
98 |
+
rank: 0
|
99 |
+
remove_short_samples: true
|
100 |
+
resume: ''
|
101 |
+
sampling_rate: 24000
|
102 |
+
save_interval_steps: 50000
|
103 |
+
stft_loss_params:
|
104 |
+
fft_sizes:
|
105 |
+
- 1024
|
106 |
+
- 2048
|
107 |
+
- 512
|
108 |
+
hop_sizes:
|
109 |
+
- 120
|
110 |
+
- 240
|
111 |
+
- 50
|
112 |
+
win_lengths:
|
113 |
+
- 600
|
114 |
+
- 1200
|
115 |
+
- 240
|
116 |
+
window: hann_window
|
117 |
+
subband_stft_loss_params:
|
118 |
+
fft_sizes:
|
119 |
+
- 384
|
120 |
+
- 683
|
121 |
+
- 171
|
122 |
+
hop_sizes:
|
123 |
+
- 30
|
124 |
+
- 60
|
125 |
+
- 10
|
126 |
+
win_lengths:
|
127 |
+
- 150
|
128 |
+
- 300
|
129 |
+
- 60
|
130 |
+
window: hann_window
|
131 |
+
train_dumpdir: dump/train_nodev_all/norm
|
132 |
+
train_feats_scp: null
|
133 |
+
train_max_steps: 1000000
|
134 |
+
train_segments: null
|
135 |
+
train_wav_scp: null
|
136 |
+
trim_frame_size: 2048
|
137 |
+
trim_hop_size: 512
|
138 |
+
trim_silence: false
|
139 |
+
trim_threshold_in_db: 60
|
140 |
+
use_feat_match_loss: false
|
141 |
+
use_subband_stft_loss: true
|
142 |
+
verbose: 1
|
143 |
+
version: 0.4.0
|
144 |
+
win_length: 1200
|
145 |
+
window: hann
|
stats.h5
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3144070fcf475d49873d9ee9c5433ae222587e569d38dd6fee8c6fff91d69c94
|
3 |
+
size 4736
|